International Journal of Mineral Processing

Transcription

1 International Journal of Mineral Processing 98 (2011) Contents lists available at ScienceDirect International Journal of Mineral Processing journal homepage: CFD modeling of the launder of settler of an industrial copper solvent extraction plant: A case study on Sarcheshmeh copper complex, Iran Roohollah Sadeghi a, Ali Mohebbi a,, Mehdi Baniasadi b a Department of Chemical Engineering, College of Engineering, Shahid Bahonar University of Kerman, Kerman, Iran b Member of Young Researchers Society, Shahid Bahonar University of Kerman, Iran article info abstract Article history: Received 28 January 2010 Received in revised form 18 July 2010 Accepted 19 October 2010 Available online 29 October 2010 Keywords: Solvent extraction Settler Launder Picket fence CFD In this study, a computational fluid dynamics (CFD) modeling accompanied by experimental field measurements has been applied to study the behavior of aqueous organic dispersion on the performance of the launder of the settler of the copper solvent extraction plant at the Sarcheshmeh copper complex, Iran. Fluid flow field has been calculated by solving the continuity and Navier Stokes equations along with the standard k-ε turbulence model. The simulation results have been compared with the field measurements data to check the accuracy of the CFD work. The effects of picket fence on the launder performance and pressure drop over the picket fences in the launder have been investigated. The model predicts that by setting the picket fences, flow pattern becomes uniform, and turbulent eddies disappear. The results also show that by installing two picket fences without dam in the launder, the phase separation is improved and the performance of the launder is optimized Elsevier B.V. All rights reserved. 1. Introduction The Sarcheshmeh copper complex deposit contains 1 billion tons of ore including 0.7% copper and 0.025% molybdenum. It is located in the southeast of Iran in Kerman province and currently processes 41,000 t/d. The Sarcheshmeh copper complex is in the central Zagros mountain range and is composed of stratified, sedimentary and volcanic rocks and faults. The Sarcheshmeh mine is one of the world's largest copper mines and the copper complex was built nearly 25 years ago. Settlers that exist in the Sarcheshmeh copper complex have launders for making uniform feed. The dispersion phase from the mixer before entering to the settler, it enters to the launder, in which a preliminary phase separation is done. 20% to 30% of phase separation occurs in the launder. Incorrect feed system in the launder results in circulation flow and macro eddies. These conditions lead to undesired phase separation. Because the phase separation in a launder is similar to a settler and there are no studies on the launder in literature, in this study, literature review was done on the settler. Solvent extraction refers to distribution of a solute between two immiscible liquid phases in contact with each other for separation of components in solution. Solvent extraction is used in numerous chemical industries to produce pure chemical compounds ranging from pharmaceuticals and biomedical to heavy organics and metals in analytical chemistry and in environmental waste purification. It has Corresponding author. addresses: amohebbi2002@yahoo.com, amohebbi@mail.uk.ac.ir (A. Mohebbi). been used more widely in hydrometallurgy (Copper, Cobalt, Nickel and Zinc). In hydrometallurgy the most applied solvent extraction method is a mixer-settler process, which fits in the group of stage wise equipment (Ritcey and Ashbrook, 1984). The main ideas in the development of the solvent extraction settler process are to achieve clean phase separation, minimize the loss of the reagents and decrease the surface area of the settlers (Lewis, 1979; Mizrahi and Barnea, 1973; Pekkala et al., 1999). In the solvent extraction process, there are two phases: aqueous and organic. Aqueous and organic phases are pumped into a mixer to achieve homogeneous dispersion. After mixing, the dispersion is fed into a settler where the aqueous and organic phases are separated by gravity. The performance of a settler depends on distribution of feed into the settler. Feed distribution has a large effect on flow pattern in the settler. Undesirable feed distribution lead to production of macro eddies and circulation flow in the settler. Macro eddies can be eliminated by baffles and packed media. The most common method to make phase separation more effective is to use picket fences, which are installed in different locations in the settler. The idea of the picket fences is to retain a deep and dense dispersion layer in the first part of the settler (Nyman et al., 1996). Several studies have been reported on the effect of picket fences in the settler. Kankaanpää (2005) confirmed that the local velocity vectors can be propagated with using at least two picket fences before impressing of viscous effect. It was also confirmed that without any picket fences the feed spouting velocity was propagated for long distances in the settler. According to Kankanpää's work (Kankaanpaa, 2005) the feed model without full depth resulted in a reverse flow of the aqueous from half way down the settler back to the feed end /$ see front matter 2010 Elsevier B.V. All rights reserved. doi: /j.minpro

2 56 R. Sadeghi et al. / International Journal of Mineral Processing 98 (2011) Table 1 Physical properties of the liquid phases (1.6 g/l Cu, ph=2 and temperature=14 C). Aqueous dynamic viscosity (mpa.s) 2.3 Organic dynamic viscosity (mpa.s) 3.3 Aqueous density (kg m 3 ) 1100 Organic density (kg m 3 ) 806 Interfacial tension (mn m 1 ) 26.1 Stanbridge and Sullivan (1999) studied inlet feed arrangement distribution to a settler. They showed that feed arrangement distribution has a great effect on the settler performance. Their study showed that plug flow in the settler can be obtained by uniform feed distribution along and width of the settler. Miller et al. (2006) reported that functionality of a settler can increase up to 30% to 50% by including better feed distribution and improved coalescence systems. These systems also increase settler performance when emulsion band is not present due to fast breaking emulsions. In the present study, a three-dimensional CFD modeling has been used to predict the performance of the launder of settler of solvent extraction plant in Sarcheshmeh copper complex, Iran. The performance of the model has been compared with the field measurement data. Moreover, the effect of picket fences on phase separation in the launder has been studied. 2. Field measurements Lack of experimental measurements for this type of launder forced us to do experimental field measurements for the current situation of Table 2 Comparison between experimental data and simulation results for aqueous phase volume fraction in the upper outlet. Sections Section 1 Section 2 Section 3 Section 4 Average through launder Experimental data Simulation results Error (%) the launder in Sarcheshmeh copper complex. The measured physical properties of the phases (organic and aqueous) are presented in Table 1. One of the important parameters in the performance of the launder of a settler is droplet size distribution of dispersed phase in the launder inlet. In the launder of Sarcheshmeh copper complex, organic phase is continuous and aqueous phase is dispersed. For measuring droplet sizes (Sinah et al., 2008), a dispersion sample (i.e. a mixture contains aqueous phase droplets dispersed in an organic phase) was poured in a Petri dish containing surfactant (Sorbitan monooleate) and organic phase. The surfactant prevented the coalescence of droplets and thus stabilized the dispersion. The dish was kept under a microscope with a camera, which is connected to a personal computer. For measuring adequate numbers of droplets, several images from different locations of the dish were taken. Fig. 1 shows a graphical droplet size distribution of dispersed aqueous phase in the launder inlet. 3. CFD simulation The Eulerian Eulerian multiphase model has been used because of high volume fraction of dispersed phase. Furthermore, the droplet coalescence and the break-up models have been implemented in the Eulerian Eulerian model of the commercial ANSYS CFX-11 software. It is assumed that the flow is turbulent, incompressible and isothermal. The calculated Reynolds number at the inlet of the launder based on hydraulic diameter was about 11, Continuity equations The continuity equations for continuous (c) and dispersed (d) phases are as follow (White, 1991; Versteeg and Malalasekera, 1995): Fig. 1. Droplet size distribution. ðα c U c Þ =0 ð1þ Fig. 2. Launder geometry.

3 R. Sadeghi et al. / International Journal of Mineral Processing 98 (2011) Source term included the buoyancy, drag, lift and turbulent dispersion forces. The buoyancy force, which is due to the density difference of the continuous and dispersed phase, could be presented in the following form for the dispersed phase (ANSYS CFX 11 User' Guide, Copyright, ): F buo d = α d ðρ d ρ 0 Þg ð6þ where ρ 0 is a reference density and g is the acceleration of gravity vector. The buoyancy force for the continuous phase is zero, (F c buo =0) because the density of the continuous phase is the buoyancy reference density. The drag force between phases can be written as: Fig. 3. Comparison between field measurements data and simulation results for different turbulence models in different sections of launder. ðα d U d Þ =0 ð2þ where α and U are volume fraction and mean velocity vector respectively Momentum equations The time average governing equation for a steady incompressible, turbulent flow form as (White, 1991; Soo, 1999; Lunder and Spalding, 1974): ρ c h i α c U c U cþ = α c P + ðμ c + μ t Þ 2 ðα c U c Þ + S M c ð3þ ρ d ½ ðα d U d U d ÞŠ = α d P + ðμ d + μ tþ 2 ðα d U d Þ + S M d ð4þ here μ, μ t, P and S M are dynamic and turbulent viscosities, pressure and source term respectively. U c and U d are the time averaged mean vectorial velocities. The k-ε turbulent model has been used to find the turbulent viscosity (Lunder and Spalding, 1974): μ t = ρc μ k 2 ε where k and ε are turbulent kinetic energy and turbulent dissipation rate respectively. ð5þ F drag = ρju rel ju rel A p C D here U rel is the relative velocity between the phases, A p is the projected area of the droplet and C D is the drag coefficient. The drag coefficient depends on the droplet diameter, the flow regime and the Reynolds number, which is given by: Re = ρ c U rel d s μ c where d s is Sauter mean diameter. In this study, Ishii and Zuber (1979) drag coefficient was used. There is a pressure distribution on the upper and lower surfaces of the droplet due to velocity gradient in the fluid. This condition makes the lift force. Lift force depends on the relative velocity and the curl of the continuous phase in the following way (ANSYS CFX 11 User' Guide, Copyright, ): F lift d = α d ρ c C L ðu C U d Þ U C ð9þ where C L is lift coefficient and was set on 0.1 according to Behzadi et al. (2004). The lift force on the continuous phase is equal but opposite in sign. Effect of turbulent fluctuation on the droplets dispersion has taken into account by turbulent dispersion force. The turbulent dispersion depends on fluctuations and volume fraction gradient of the continuous phase. The dispersion force for continuous phase can be written as follows (Soo, 1999): F td c = C TD ρ c k αc αc ð10þ ð7þ ð8þ Fig. 4. Flow pattern (velocity vectors) in the launder.

4 58 R. Sadeghi et al. / International Journal of Mineral Processing 98 (2011) Fig. 5. Different planes for displaying contours of the organic phase volume fraction. where C TD is the turbulent dispersion coefficient that was set on 0.1 according to Olmos et al. (2001). k αc is the turbulent kinetic energy of the continuous phase. For the dispersed phase, this force is equal but opposite in sign. In the launder, the dispersion does not contain just one droplet size. A droplet size distribution can be used for dividing the droplet sizes into several groups. Population balance modeling gives a very good opportunity to handle the dispersed phase and its droplet size distribution. In this work, the population balance model, which is developed by Lo (Lo, 1996), was used. This model is multiple-size-group (MUSIG). The model has been incorporated into CFX 11 software and can handle dispersed multiphase flows in which the dispersed phase has a large variation in size. MUSIG provides a framework in which the population balance method together with the break-up (Luo and Svendson, 1996) and coalescence (Prince and Blanch, 1999) modelsis incorporated into three-dimensional CFD calculations. Continuity equation for one MUSIG-size group-i can be written as (Kankaanpää, 2007): h i ρ d α d; i U d = S i ð11þ where S i is the source term of the rate of mass transfer into the MUSIG-size group due to the break-up and coalescence processes. Eq. (11) can be applied to all MUSIG-size groups because the sum of all droplet volume fractions equals the volume fraction of dispersed phase, i.e.: i α d; i = αd ð12þ The individual volume fraction of MUSIG-size-group-i can be written as: α d; i = α d f i ð13þ where f i is the fraction of dispersed phase volume fraction in MUSIGsize group-i. When Eq. (13) is substituted into Eq. (11), the continuity equation for the MUSIG-size group-i is yielded: ρ d ½ ðα d f i U d ÞŠ = S i ð14þ The source term S i can be calculated from population balance (see Eq. (16)) when the product of the droplet number density of MUSIGsize group-i (n i ) and the droplet volume of MUSIG-size group-i (v i ) are related to the volume fraction of MUSIG-size group-i by (Kankaanpää, 2007): n i v i = α d f i ð15þ In this way, the normal population balance equation can be related to the continuity equation of the dispersed phase. Eq. (16) became Fig. 6. Volume fraction contours of the organic phase in the launder.

5 R. Sadeghi et al. / International Journal of Mineral Processing 98 (2011) Fig. 7. Picket fences geometry. identical to Eq. (14) when the dispersed phase density multiplies both sides (Kankaanpää, 2007). ðu d n i Þ = ðα d f i U d Þ = s 0 i = B B D B + B C D C ð16þ Where s i ' is source term of the rate of mass transfer into the MUSIG-size groupduetobreak-upandcoalescence processes with unit of 1/(m 3 s). In Eq. (16), B B,B C,D B,andD C represent the birth rate due to break-up of larger droplets, the birth rate due to coalescence of smaller droplets, the death rate due to break-up into smaller droplets, and the death rate due to coalescence with other droplets respectively. During the normal iteration, additional transport Eq. (14) was solved for the scalar variablef i,afterthat,thesizedistributionofthedispersed phase can be defined from the solution of f i. Consequently, the Sature mean diameter (d s ) can be calculated from the equation in below: d s = 1 ð17þ f i i d i Fig. 8. Flow pattern (velocity vectors) in the launder with one picket fence in the location of 5.68 m from the launder inlet.

6 60 R. Sadeghi et al. / International Journal of Mineral Processing 98 (2011) Fig. 9. Volume fraction contours of the organic phase in the launder with one picket fence in the location of 5.68 m from the launder inlet. where d i is the droplet diameter of the MUSIG-size group-i. This mean diameter is used in computing the drag force between the continuous phase and dispersed phase in momentum equations. 4. Physical domain Fig. 2 shows the launder geometry of the Sarcheshmeh copper complex, which has been used in this study. Its dimensions are 17.8 m 8.9 m 0.3 m with two outlets at the end of the launder, positioned on the top and bottom of the end wall. The launder has four sections with a dam for distributing flow into the sections. Each section has a width of 2.15 m. Flow rate in the launder is 1100 m 3 /h and organic to aqueous flowrate ratio is In the launder, the organic phase is continuous and the aqueous phase is dispersed. The fluid height in the launder is 0.26 m. In Fig. 2, the x axis is the flow direction, z is vertical to the flow direction in a horizontal plane and y is the vertical axis. 5. Calculations procedure The launder process was studied in steady state. The governing mass and momentum equations were solved by using commercial ANSYS CFX-11 software package. ANSYS CFX is a commercial multipurpose CFD code currently developed by ANSYS Inc. The CFX numerical kernel uses the element based finite volume method (EbFVm) to treat generalized unstructured meshes in Cartesian coordinates. The discrete system of linearized equations is solved using the algebraic multigrid (AMG) method accelerated by the incomplete lower upper (ILU) factorization technique. The pressure velocity coupling is carried out in a single cell of the collocated grid using a Rhie and Chow (Rhie and Chow, 1983) like formulation. This Table 3 Aqueous phase volume fraction in the upper outlet and pressure drop over picket fences. Aqueous phase volume fraction Pressure drop (Pa) One picket fence Two picket fences One picket fence without dam Two picket fences without dam Pressure drop (cm dispersion) solution approach uses a fully implicit discretization of the equations. In steady state solutions, the false time step technique is applied to the solution relaxation. UPWIND differencing scheme was employed for volume fraction and the High Resolution scheme used for the other terms. The mass flow, hydrostatic pressure, frictionless and no-slip boundary conditions at the inlet, outlet, free fluid surface and walls have been applied respectively. Moreover, direction of flow in the inlets was chosen according to actual flow directions in the launder. The grid independency test was carried out with three different three-dimensional meshes, which contained 300,000, 400,000 and 500,000 cells respectively. Comparison between simulation results and experimental data showed that 400,000 cells were needed to produce the grid-independent solution. Several parameters must be selected before calculations in MUSIG model. These parameters are number of the droplet size groups, the minimum droplet diameter, the maximum droplet diameter, and the coalescence and break-up calibration coefficients. Fig. 1 shows that minimum droplet size in the inlet is about 100 μm. Therefore,the minimum size in the MUSIG model was selected 100 μm. The coalescence calibration coefficients of 0.01, and were used in the different simulation runs with maximum droplet size of 1200 μm and number of the droplet size groups of 10. When the coalescence calibration coefficient was 0.1, effective coalescence occurred in the entrance of settler and the separated aqueous phase contained only with maximum size group. When the coalescence calibration coefficient was 0.005, the droplet size distribution became uniform but the amount of aqueous phase in the upper outlet was much. Weak phase separation occurred in the settler for coalescence calibration coefficient of because the separation of phases in the settler was controlled by the coalescence rather than break-up; therefore, the calibration coefficient of break-up did not affect on the process. Different simulation runs were done for determining maximum droplet size. The maximum droplet sizes of 1000, 1200, 1400, 1600, 1800, 2000 and 2200 μm were used. The results of simulations showed that with increasing maximum droplet size, proper droplet size distribution in different regions in the settler obtained. Different simulation runs were carried out with number of droplet size groups of 10, 15 and 20. When the number of droplet size groups increased, increasing droplet size from minimum to maximum values could be smoother and it was possible that droplet sizes in the inlet of settler were distributed in the adequate size groups according to the measured droplet distribution in Fig. 1. On the other hand as the numbers of size groups increase, the calculation time increases. After these test calculations, for effective phase separation, the coalescence and break-up

7 R. Sadeghi et al. / International Journal of Mineral Processing 98 (2011) Fig. 10. Flow pattern (velocity vectors) in the launder with one picket fence in distance in the location of 10 m from the launder inlet. calibration coefficients, maximum droplet size and number of droplet size groups were selected 0.005, 0.1, 2000 μm and 15 respectively. 6. Results and discussion 6.1. Validation of the model Measured data from Sarcheshmeh copper complex, Iran were used for validation of modeling. A comparison between the field measurements and the modeling for volume fraction of aqueous phase in the upper outlet, which is a criterion for separation efficiency, is given in Table 2. Flow field was obtained by standard k-ε model. The simulation results are in good agreement with the field measurements. The flow field was also calculated by RNG k-ε and k-ω turbulence models in addition to standard k-ε model. Fig. 3 shows a comparison between field measurements and simulation results for different turbulence models. It can be seen that there are a little difference between the models. Therefore, the k-ε turbulence model was applied in this work because the computational time used in this model was much smaller than that for other models. Fig. 4 shows flow pattern (velocity vectors) in the launder. The CFD shows that there is a significant asymmetry in the flow with recirculation zones forming in the sections 2 and 4 of the launder. It seems reasonable to suppose that presence of these zones is causing short circuiting of the flow, with the poorer separation in these sections being attributable to this effect.however, this asymmetry in the CFD predictions has been to a large degree induced because of flow asymmetry at the inlets. This is clearly shown with the vectors on the upper inlet in Fig. 4 parallel to the launder axis and the vectors on the lower inlet normal to the launder axis. Asymmetry in the flow in the launder generates recirculation zone and macro eddies, which leads to increase turbulent mixing and lift force. This causes increasing break-up and decreasing coalescence. Therefore, these aspects contribute to the poor separation in the Fig. 11. Volume fraction contours of the organic phase in the launder with one picket fence in the location of 10 m from the launder inlet.

8 62 R. Sadeghi et al. / International Journal of Mineral Processing 98 (2011) Fig. 12. Flow pattern (velocity vectors) in the launder with two picket fences. launder. The macro eddies that are inherent in the launder can be eliminated by either eliminating the cause of the eddy, or by dissipating the eddy flow pattern by use of picket fences (see Section 6.2). Fig. 5 shows different planes for displaying organic phase volume fraction in the launder. Fig. 6 shows volume fraction contours of the organic phase. This figure shows that phase separation is not good. In this case, volume fraction of the aqueous phase in the upper outlet is Effect of picket fences on launder performance Increasing the performance of the existing launder in Sarcheshmeh copper complex has been the main goal of this research. To achieve this goal the launder operation has been optimized by testing different arrangements of picket fences with the aid of CFD modeling. Picket fences geometry in two rows is shown in Fig. 7. The dimensions of each picket fence are 0.25 m 0.3 m m and the distance between two picket fences in each row is m. The effect, location and number of picket fence rows on the performance of the launder have been the case studies. The operating conditions for simulation of the launder with and without picket fence row were the same. The effect of location of picket fences on the launder performance has been tested by installing the picket fence row at two locations. Flow patterns (velocity vectors) in the launder with one picket fence row are shown in Fig. 8. The distance of this row from the inlet of the launder was 5.68 m, which was the closest possible distance allowed for installation of the picket fences. It can be seen from Fig. 8 that by installing this picket fence row, the flow becomes uniform. Fig. 9 shows volume fraction contours of the organic phase in the launder with one picket fence. This figure shows as a result of installation of the picket fence, the phase separation is more than the case of without picket fence. In this case, volume fraction of the Fig. 13. Volume fraction contours of the organic phase in the launder with two picket fences.

9 R. Sadeghi et al. / International Journal of Mineral Processing 98 (2011) Fig. 14. Flow pattern (velocity vectors) in the launder with one picket fence and without dam. aqueous phase in the upper outlet reaches to instead of 0.26 for the case of without picket fence. Also, the pressure drop over the picket fence is 6.7 cm (see Table 3). The effect of installation of picket fence on the launder performance in another location was investigated. In this case, the picket fence was installed in the location of 10 m from the inlet instead of 5.68 m. The results were shown in Fig. 10. Ascanbeseen in this figure the flow pattern in the launder is not uniform and there are circulation flows. Fig. 11 shows the volume fraction contours of the organic phase for this case. This figure shows that the phase separation is not good and the picket fence do not affect on separation of phase layers. As a new case, the effect of installation of two picket fence rows in the location of 5.68 m from the launder inlet was investigated. Fig. 12 shows flow pattern (velocity vectors) with two picket fence rows. As one can see in this figure, the flow is completely uniform without circulation zone. The volume fraction contours of the organic phase in the launder with two picket fence rows are shown in Fig. 13. This figure shows that phase separation in the launder is done better than that of one picket fence. In this case, according to Table 3 the aqueous phase volume fraction in the upper outlet and pressure drop over the picket fences are 0.18 and 11 cm respectively. Fig. 13 also shows existence of the dam causes to disturb separation of phases and to increase pressure drop. This results in overflow in the launder. Therefore, simulation without dam was done. Flow pattern (velocity vectors) and the organic phase volume fraction contours in the launder with one picket fence without dam are shown in Fig. 15. Volume fraction contours of the organic phase in the launder with one picket fence and without dam.

10 64 R. Sadeghi et al. / International Journal of Mineral Processing 98 (2011) Fig. 16. Flow pattern (velocity vectors) in the launder with two picket fences and without dam. Figs. 14 and 15 respectively. As one can see in these figures separation phases are not good and the flow in the launder is circulation. The volume fraction of the aqueous phase in the upper outlet reaches 0.42 and the pressure drop over the picket fence is 3 cm (see Table 3). As a final case, the effect of installation of two picket fence rows in the location of 5.68 m from the inlet without dam was investigated. Figs. 16 and 17 show flow pattern (velocity vectors) and the volume fraction contours of the organic phase in the launder with two picket fence rows and without dam respectively. In this case, flow in the launder is almost uniform and without circulation and phase separation is good. As one can see in Table 3, whenthere aretwopicketfencerowswithoutdam,thevolumefractionofthe aqueous phase in the upper outlet and pressure drop over the picket fences are 0.17 and 9 cm respectively. This is the best case for optimizing the performance of the launder. 7. Conclusions In this study, a three-dimensional computational fluid dynamics has been applied to predict the performance of an industrial settler's launder in Sarcheshmeh copper complex. The effect of picket fences on phase separation in the launder and the factors affecting coalescence were discussed. The simulation results were validated by the experimental data and showed that by setting the picket fences, flow pattern becomes uniform and recirculation region disappears. Therefore, phase separation becomes more effective and the entrainment amounts in the separated phases can be decreased. By increasing the number of picket fences, phase separation and pressure drop increase. From the results of simulation, it can also be concluded that if two picket fences without dam are set, the phase separation will improve and the pressure drop has a suitable value. Finally, from the results of this work, it can be concluded that Fig. 17. Volume fraction contours of the organic phase in the launder with two picket fences and without dam.

11 R. Sadeghi et al. / International Journal of Mineral Processing 98 (2011) computational fluid dynamics is a powerful tool for simulation and optimization of industrial launders in a solvent extraction plant without paying additional cost. Acknowledgment The authors would like to acknowledge Sarcheshmeh copper complex for their financial support. References Ritcey, G.M., Ashbrook, A.W., Solvent Extraction, Principles and Application to Process Metallurgy, Part I. Elsevier Science Publisher B.V., Amsterdam. Lewis, I.E., Design of mixer-settler to achieve low entrainment losses and reduce capital costs, ISEC 1977, Toronto, CIM Special Volume, 21, pp Mizrahi, J., Barnea, E., Compact settler give efficient separation of liquid/liquid dispersions. Process Engineering 1, Pekkala, P., Kuusisto, R., Lyyra, J., Nyman, B., Lindell, E., Ekman, E., October 10 13, Solvent Extraction How to Get Over Hard Times. In: Young, S.K., Dreisinger, D.B., Hackl, R.P., Dixon, D.G. (Eds.), Proceedings of the Copper 99 Cobre 99, the 4th International conference, Vol. IV(Hydrometallurgy of copper). The Minerals, Metals & Materials Society, Phoenix, USA, pp Nyman, B., Kuusisto, R., Taipale, P., Lyyra, J Emphasis on feed end settling in Outokumu's copper VSF mixer-settler, ALTA 1996 Copper Hydrometallurgy Form, Brisbane, Australia, October 9, Kankaanpaa, T., February 13 17, In: Schlesinger, M.E. (Ed.), Studying Solvent Extraction Settler Process by Using CFD, EPD Congress. TMS (The Minerals, Metals & Materials Society), San Francisco. Stanbridge, D., Sullivan, J One example of how offshore gil & gas industry technology can be of benefit to hydrometallurgy, second international conference on CFD in the minerals and process industries, CSIRO, Melbourne, December. Miller, G Design of mixer-settlers to maximize performance, Miller Metallurgical Services, ALTA COPPER-10. Sinah, K.K., Mahajan, S.M., Shonog, K.T., Ghosh, S.K., Representative drop size and drop size distribution in A/O dispersion in continuous flow stirred tank. Hydrometallurgy 90, White, F.M., Viscous Fluid Flow, 2nd ed. McGraw Hill, Singapore. Versteeg, H.K., Malalasekera, W., An Introduction to Computational Fluid Dynamics, The Finite Volume Method. Longman Group Ltd., New York. Soo, L.S., Multiphase Fluid Dynamics. Science Press, Hong Kong. Lunder, B.E., Spalding, D.B., The numerical computation of turbulent flows. Computer Method in Applied Mechanics and Engineering 3 (2), ANSYS CFX 11 User' Guide, Copyright , ANSYS, Ltd. Ishii, M., Zuber, N., Drag coefficient and relative velocity in bubbly, droplet or particulate flows. AIChE Journal 25 (5), Behzadi, A., Issa, R.I., Rushe, H., Modelling of dispersed bubble and droplet flow at high phase fractions. Chemical Engineering Science 59, Olmos, E., Gentric, C., Vial, Ch., Wild, G., Midoux, N., Numerical simulation of multiphase flow in bubble column reactor. Influence of bubble coalescence and break up. Chemical Engineering Science 56, Lo, S.M Application of the MUSIG model to bubbly flows, AEAT Technology. Luo, H., Svendson, H.F., Theoretical model for drop and bubble break up in turbulent dispersions. AIChE Journal 42 (5), Prince, M.J., Blanch, H.W., Bubble coalescence and break up in air sparged bubble columns. AIChE Journal 36 (10), Kankaanpää, T CFD Procedure for Studying Dispersion Flows and Design Optimization of the Solvent Extraction Settler, Doctoral Thesis, Helsinki University of Technology, Finland. Rhie, C.M., Chow, W.L., Numerical study of the turbulent flow past an airfoil with trailing edge separation. AIAA Journal 21,